Krutrim, India's first generative AI unicorn, is abandoning its ambition to build frontier AI models and pivoting to cloud services. The startup — founded by Ola CEO Bhavish Aggarwal — has laid off over 200 employees, pulled its AI assistant app from stores, paused chip design efforts, and gone silent on product updates for months. The pivot reflects a harsh reality: building competitive AI models requires resources that only the world's largest companies can sustain.
What Happened to Krutrim
Krutrim launched with enormous ambition. In January 2024, it raised $50 million at a $1 billion valuation — becoming India's first GenAI unicorn. The company positioned itself as a domestic alternative to OpenAI, Anthropic, and Google, with plans to build large language models optimized for Indian languages.
The reality proved harder. Krutrim released its Krutrim-2 base model in February 2025. Then it went quiet. Its last post on X was in December. It skipped India's AI Impact Summit entirely. It pulled its Kruti AI assistant app from both app stores in April. And over three rounds, more than 200 employees were laid off.
The company now says it generated about $31.5 million in revenue in FY2026 — a threefold increase. But roughly 90 percent of its previous year's revenue came from Ola's own ecosystem. Whether the cloud pivot can attract genuine external customers remains an open question.
Why the Model Dream Failed
Building frontier AI models costs billions. OpenAI raised $122 billion in a single round. Anthropic is raising $50 billion at a $900 billion valuation. Google is spending $40 billion on Anthropic alone. Even well-funded European alternatives like Cohere and Aleph Alpha needed to merge to stay competitive.
Krutrim's $50 million was a rounding error in this context. India's entire startup ecosystem attracted just $74 million in AI funding over the past year. The gap between Indian AI ambitions and the capital required to compete at the frontier is enormous — a point the Cursor-Pakistan story illustrated from a different angle.
India has strong engineering talent. It has a massive and growing app market. But it does not yet have the capital infrastructure to fund frontier AI model development at the scale required to compete with American and Chinese labs.
The Cloud Pivot Makes Sense
Krutrim says it now has over 25 enterprise customers across telecom, financial services, and healthcare. Most of its GPU compute capacity is committed to external workloads. The cloud business is generating revenue and margins above 10 percent.
The pivot follows a familiar pattern. Building AI infrastructure — selling compute, storage, and deployment services — is a more viable near-term business than building frontier models. It is also what India's market needs. Indian companies want to deploy AI. They do not necessarily need homegrown models to do it. Google, OpenAI, and Anthropic's models work fine. What they need is affordable, locally compliant cloud infrastructure to run them on.
Sarvam Keeps Going
While Krutrim retreated, rival Indian AI lab Sarvam has continued to release new models, sign partnerships, and participate in major events. Sarvam showcased open-source models, hardware developments, and a deal with space-tech firm Pixxel at India's AI Impact Summit.
The contrast is stark. Both companies were positioned as India's answer to the global AI race. Sarvam has continued executing. Krutrim has pivoted away from its original mission. Whether Sarvam can sustain its trajectory with the limited capital available in India's market is its own question. But for now, it remains the more credible contender.
What It Means
Krutrim's pivot is a cautionary tale for AI startups in emerging markets. The aspiration to build sovereign AI models is understandable. The economics make it nearly impossible without access to billions in capital. In a world where frontier models cost tens of billions to train and the GPU shortage is squeezing everyone from startups to universities, a $50 million unicorn cannot compete on model development.
The smarter play — as Krutrim now acknowledges — may be to build the infrastructure that helps local companies deploy the models someone else already built. It is a smaller ambition. But it is a viable business. And in India's AI market, viability matters more than dreams.
